Re: [boost] [histogram] review part 2 (implementation)

On 26. Sep 2018, at 10:41, degski wrote:

On Wed, 26 Sep 2018 at 11:17, Hans Dembinski via Boost mailto:boost@lists.boost.org> wrote: Ok, no calling of destructors.

You should, iff you use placement new to construct an object in uninitialized memory, that's what Steven started with saying.

Right, I got this now. When you have a union type with non-trivial bounded types you are also supposed to call the destructor manually when you switch types, and there it makes sense. My mistake was to believe that calling the destructor manually is like calling any other function. Now it is clear.